It uses an IV approach to estimate a causal effect; the description of the IV is:
> [The study] exploits the authors’ novel observation that, in the UK, the timing of birth affects breastfeeding for low educated mothers. In particular, amongst this group of mothers, breastfeeding rates are lower for those who give birth just before or early into the weekend compared to those who give birth at any other time during the week. We argue that this is because the provision of infant feeding support in UK hospitals is lower at weekends than during the week. Without early hands-on support at the hospital, it is much more difficult for successful breastfeeding to be established. At the same time, we provide extensive evidence that maternal and birth-related characteristics do not vary by timing of birth, and nor do a range of other hospital maternity services vary by timing of birth. Timing of delivery therefore provides a credible source of exogenous variation that we use as an instrumental variable for breastfeeding.
Using that IV they find "large effects of breastfeeding on children's cognitive development but no effects on health or noncognitive development during the period of childhood we consider."
I had seen this one! I don't really know what to make of it - the effects are absolutely enormous, and the confidence intervals around them are massive too. I know they do a bit to address that in the extra analyses in the study. But... nearly ten whole IQ points!? They say it's "not implausible" given the effects in the RCT, but given the effects in the RCT aren't very plausible either, I'm not sure how much of a triangulation that is...!
Large confidence intervals, but they do report p=0.0078 for cognitive development, which suggests that there is a positive effect on cognition even in developed countries, no? If you think the study is flawed I’d be v. keen to hear more -- I tend to take causal designs seriously even when they are not RCTs.
Can I as, why you find the effect size implausible? Is it because you think the fact that size should be visible by eye? (The only thing you’ve said in all 8 eps that I’d really push back on is that we should be able to see moderate effects by eye! I have some counterexamples to that, which I can email you if you are interested.)
Glad to have the record corrected, let the public know that Stuart likes Coldplay.
No no no no no no no no no no no
Really enjoying the show! I emailed you about a recent study on breastfeeding, but it looks like it didn't reach you...
Fitzsimons & Vera-Hernández (2022). Breastfeeding and Child Development
https://www.aeaweb.org/articles?id=10.1257/app.20180385
It uses an IV approach to estimate a causal effect; the description of the IV is:
> [The study] exploits the authors’ novel observation that, in the UK, the timing of birth affects breastfeeding for low educated mothers. In particular, amongst this group of mothers, breastfeeding rates are lower for those who give birth just before or early into the weekend compared to those who give birth at any other time during the week. We argue that this is because the provision of infant feeding support in UK hospitals is lower at weekends than during the week. Without early hands-on support at the hospital, it is much more difficult for successful breastfeeding to be established. At the same time, we provide extensive evidence that maternal and birth-related characteristics do not vary by timing of birth, and nor do a range of other hospital maternity services vary by timing of birth. Timing of delivery therefore provides a credible source of exogenous variation that we use as an instrumental variable for breastfeeding.
Using that IV they find "large effects of breastfeeding on children's cognitive development but no effects on health or noncognitive development during the period of childhood we consider."
I had seen this one! I don't really know what to make of it - the effects are absolutely enormous, and the confidence intervals around them are massive too. I know they do a bit to address that in the extra analyses in the study. But... nearly ten whole IQ points!? They say it's "not implausible" given the effects in the RCT, but given the effects in the RCT aren't very plausible either, I'm not sure how much of a triangulation that is...!
Large confidence intervals, but they do report p=0.0078 for cognitive development, which suggests that there is a positive effect on cognition even in developed countries, no? If you think the study is flawed I’d be v. keen to hear more -- I tend to take causal designs seriously even when they are not RCTs.
Can I as, why you find the effect size implausible? Is it because you think the fact that size should be visible by eye? (The only thing you’ve said in all 8 eps that I’d really push back on is that we should be able to see moderate effects by eye! I have some counterexamples to that, which I can email you if you are interested.)
Also see https://neurosciencenews.com/breast-milk-alternative-iq-23851/ which is consistent with the idea that there is a substantial gap (but that it’s closing as formula improves).